Conversation
Setting Format on a Run Pandoc Export step that also has a preset had no effect, and nothing said so: the only signal was a console.warn. The precedence itself is deliberate and documented — this makes it visible. - The step editor greys the Format control out while a preset is set and says which preset decided the format. Declared as a generic `disabledBy` / `disabledDescription` pair on CompileStepOption, resolved by one shared helper both option renderers now use. - The Template / preset dropdown labels each preset with what it produces (`paperbell — PDF`, `manuscript-obsidian — DOCX`), read from the preset's own yaml via the existing exportTargetForDefaults. Best-effort: an unreadable preset keeps its bare name. - A Notice on export covers the paths the editor can't: workflows saved before this, and the headless `Run workflow: <name>` command. Closes #38 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_013ivDqLS8ZQvs3gVQ7qeJGC
The verification owed by #25 that unit tests could not close. Two halves, because the workflows differ: - Manuscript / SI / Response open with Strip Frontmatter, which drops the whole yaml block — asserted directly on the step, for both scene and manuscript kinds. It had no test file of its own until now. - Cover Letter has no such step: its single Run Pandoc Export hands the note's frontmatter to pandoc on purpose, since the cover_letter template reads to/date/manuscript from it. So that one gets a real pandoc run, gated on the assets like the response-letter golden, asserting the template does read the frontmatter and that `project:` still emits nothing. Refs #28 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_013ivDqLS8ZQvs3gVQ7qeJGC
The `paperbell:plugins-changed` event constant was vendored but never subscribed. It is the only consent-free push the host gives us between ready events, so take it as a cue to re-read plugin-info: capabilities the host gained or dropped mid-session no longer sit stale until it reloads. Registers nothing and prompts for nothing. The capabilities read moves out of `attach()` into a small method both paths share. Refs #33 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_013ivDqLS8ZQvs3gVQ7qeJGC
Every new paper started by hand-correcting four fields the PaperBell host already knows. Its v2 `profile` now fills `creators[0]` in both metadata.json files and the cover letter's `corresponding:` line, each field falling back to its own placeholder — no host, an older host, a half-filled profile and an ungranted scope all still produce today's scaffold byte for byte. `profileIfGranted()` reads the profile only when that costs no consent prompt: the config the host already pushed, or a fetch when listGrants() says `config` is already granted. A dialog raised by opening the new-paper modal could outlive it (#31), and a pre-filled author is not worth that. The name is written exactly as the host gave it — it is a display name and metadata.json's placeholder follows Zenodo's "Lastname, Firstname", but reordering it would be a guess. ORCID stays a placeholder; the host has no such field. Closes #30 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_013ivDqLS8ZQvs3gVQ7qeJGC
The release process documented the release PR but said nothing about how the feature PR before it should be merged — which is what let a merge commit inheriting a `feat:` PR title double-count in 2.4.0-beta.5's changelog. Squash, and say why. Closes #34 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_013ivDqLS8ZQvs3gVQ7qeJGC
Follow-up to the review of this branch.
- Split the preset listing along the seam MAINTAINING.md documents: the
fs + parseYaml half stays in pandoc-templates.ts, the pairing and its
degradation move to a pandoc-templates-utils.ts sibling that vitest can
load — so "an unreadable preset keeps its name and loses its label" is
now asserted rather than asserted-to-be-obvious.
- The step editor's disabled description now names the format the preset
imposes ("decides the output format (PDF)"), which is what #38 asked
for. The export-time warning and notice are filled from that same
string via `fillOptionText`, so the three copies of the sentence become
one declaration and three consumers. The notice moves after the target
is resolved, which is where the format is known.
- `leadCreator(ctx)` replaces the creators[0] block duplicated between
the main and supplementary metadata.
- `listPandocTemplateNames` for the error path that only wants names;
`dynamicChoices` narrowed to its one literal; assorted one-liners
(discarded return, over-defensive branches, mixed ctx access).
- `hasBin`/`ASSETS` for the golden tests extracted to a shared harness
rather than copied a third time.
Left alone deliberately: the notice fires on every export (a saved
misconfiguration is worth repeating), compile-step strings stay English
(file-local precedent), and ScaffoldProfile keeps copying fields rather
than aliasing the vendored contract type — that decoupling is the point.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013ivDqLS8ZQvs3gVQ7qeJGC
…gnal fix: make the pandoc preset/Format precedence visible, plus four small issues
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Release candidate for 2.4.0-beta.7. Everything since
2.4.0-beta.6, from PR #39.Run Pandoc Exportgreys out Format while a preset is set, naming the preset and the format it produces (Ignored — the preset "paperbell" decides the output format (PDF).). The Template / preset dropdown labels every entry —paperbell — PDF,manuscript-obsidian — DOCX. A notice covers saved workflows and the headlessRun workflow:path. The precedence itself is unchanged.creators[0]and the cover letter'scorresponding:from the host's v2profile, per field, and never at the cost of a consent prompt. No host / older host / ungranted scope → today's scaffold byte for byte.MAINTAINING.md— and this PR is the first one to follow it: squash-merge, please.project:key verification, as tests: a Strip Frontmatter unit test, plus a real-pandoc golden for the Cover Letter workflow (the one that keeps the note's frontmatter on purpose). Manual steps still open on the issue.paperbell:plugins-changedand re-read capabilities. The read API / event bus stays open, blocked on a host decision.Plus a quality pass over the above (
008c02f): the preset listing split along the*-utils.tsseamMAINTAINING.mddocuments and covered by tests, the precedence sentence reduced to one declaration with three consumers,leadCreator(ctx)deduplicated across bothmetadata.jsonfiles, and the golden tests'hasBin/ASSETSshared instead of copied a third time.Why this merge should be squashed
2.4.0-beta.6's changelog carries the same fix twice — once from PR #36's merge commit, once from the original — which is exactly what #34 diagnoses and what this PR documents the fix for.Verification
npm run build(svelte-check + eslint + rollup) clean;npm run test:unit485 passed / 2 skipped, including the cover-letter and response-letter goldens, which really run against the local pandoc +pandoc-assets/.Closes #38
Closes #30
Closes #34
🤖 Generated with Claude Code
https://claude.ai/code/session_013ivDqLS8ZQvs3gVQ7qeJGC